# Random Design [Health: Active]

**Category:** 📐 Architecture & Design  
**Repository:** https://github.com/yanislav-igonin/random-design-mcp  
**GitHub Stars:** 0  
**npm Downloads (last month):** 133  
**Views:** 0  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/random-design

## Description
Generate randomized English Markdown frontend design directions from curated catalogs.

## Tools
Capabilities this server exposes over MCP:

- **productType** — what you are designing, such as `landing page`, `dashboard`,
- **audience** — who the interface is for, such as `developers`, `founders`, or
- **priority** — what the design should optimize for or pay extra attention to,
- **compatibility** — controls how wild the random combination can be. It defaults

## Claude Desktop Quick Installation
Install path detected from listing signals. Uses `npx` (confidence: high):

```json
"mcpServers": {
  "random-design": {
    "command": "npx",
    "args": ["-y","random-design-mcp"]
  }
}
```

## Documentation

## What the Random Design MCP server does

The Random Design MCP server returns one randomized frontend design direction as English Markdown. Each result is assembled from curated catalogs rather than generated by an LLM, making the server suitable for producing structured starting points for interface exploration.

You can leave all inputs empty for an unconstrained request, or provide context to make the direction more relevant. Supported context includes the kind of product being designed, its intended audience, and the design priority. Examples include landing pages, dashboards, mobile apps, pricing pages, developers, founders, conversion, readability, premium feel, dense data display, and playful experimentation.

The Random Design MCP server is intended for ideation rather than implementation.

## How it works

The package exposes MCP tools through a standard stdio process. An MCP client starts the published npm package with `npx`, sends a tool request, and receives Markdown in response. The server chooses randomized values from its catalogs and can use soft compatibility scoring to keep the combination more coherent.

The `compatibility` input defaults to `true`. With that setting, generated parameters are constrained toward combinations described as practical and compatible. Set it to `false` when you want deliberately unconstrained results that may mix clashing eras, styles, palettes, layouts, and materials.

The `generate_design_description` tool accepts four optional fields: `productType`, `audience`, `priority`, and `compatibility`. The separate `get_version` tool takes no arguments and reports the package version from `package.json`, which can help verify the version a client is running.

## Setup and configuration

Install the published package through an MCP client that supports stdio servers. A client configuration can invoke `npx` with `-y random-design-mcp`, using `random-design` as the server key. No required environment variables are documented for normal operation.

For local development, the repository provides `npm install`, `npm run build`, and `npm run check`. Run `npm run dev` to preview a generated direction locally. The development preview can read `RANDOM_DESIGN_PRODUCT_TYPE`, `RANDOM_DESIGN_AUDIENCE`, `RANDOM_DESIGN_PRIORITY`, and `RANDOM_DESIGN_COMPATIBILITY` for optional context; these variables are shown for preview configuration, not as required runtime settings.

## Tools and capabilities

- `generate_design_description`: returns one randomized Markdown design direction and accepts optional design context.
- `get_version`: returns the package version from `package.json` without requiring arguments.
- Compatibility control: keeps combinations more coherent by default or permits intentionally clashing combinations when disabled.

The Random Design MCP server does not call an LLM. Its output is based on the bundled curated catalogs and the supplied request fields.

## Limitations and notes

Results are randomized, so repeated requests can produce different directions. Compatibility scoring is soft rather than a guarantee that every result will suit a particular product. Disabling it intentionally allows combinations that may be impractical or visually conflicting.

The documented output is English Markdown frontend design guidance. Client use requires an MCP application with stdio server support.

_Full upstream README: https://allmcps.com/mcp/random-design/readme_